The ability to apply artificial intelligence to telematics data in order to predict vehicle and engine maintenance needs is maturing, showing new promise in helping fleets avoid breakdowns that lead to costly downtime.
Telematics has played a pivotal role in preventive maintenance for years, supplying data on vehicle mileage, engine hours and other variables to aid fleets in scheduling maintenance activities. This information continues to be instrumental in preserving warranty coverage that can keep repair costs down and limit insurance claims, but it is missing a key capability that is equally valuable in protecting the bottom line: predictive maintenance abilities that can provide early warnings of component failures that may cause unplanned vehicle or engine downtime.
Artificial intelligence began closing that gap several years ago and continues to expand its value as a tool for keeping breakdowns at bay. Data scientists are developing increasingly sophisticated predictive AI models that can not only analyze more inputs and their interactions with each other, but also utilize machine learning to continually improve prediction accuracy.
The goal is to identify patterns that indicate a particular system or component is operating out of normal range. The problem can then be addressed before a failure puts the truck out of service, causing missed delivery deadlines and the downstream domino effects that are every fleet manager’s nightmare.
Early Warning System
Consider the example of pump failure. Pre-AI predictive maintenance efforts utilized fixed mathematical models that monitored and set acceptable thresholds on the vibration frequency of pump bearings based on telematics data, diagnostic trouble codes and mechanics’ observations. Those thresholds were used to set general maintenance intervals that applied to all pump types as well as all vehicle models.
AI improves the ability to predict catastrophic pump malfunctions by performing a broader historical analysis spanning pump bearings, oil temperature, oil quality, fluid pressure, and in some cases maintenance logs and technician notes. The resulting model is then continually refined via machine learning to identify new patterns indicating that pumps are headed for trouble. Fleets can use these insights to fine-tune their maintenance schedules to reduce the risk of vehicle stalls, engine overheating and other pump-related emergencies.
The Internet of Things (IoT) is also contributing to advances in predictive maintenance. With data collection extended to every connected vehicle, for example, data scientists have the ability to employ a larger training set to build AI models that can even more accurately detect potential problems before they occur. In the not too distant future, that same connectivity will enable critical issues on individual vehicles to be flagged in real time and sent to the cloud for instant visibility and proactive repairs, minimizing the need for emergency maintenance and associated downtime.
On the Cutting Edge
Telematics service providers (TSPs) are beginning to offer AI-driven predictive maintenance as an optional add-on to their existing solutions, providing weekly lists of vehicles at risk of near-term breakdowns for at-a-glance review by maintenance managers.
The current challenge is that there is no one-size-fits-all predictive maintenance solution. AI predictive models need to be customized to each truck make and model because every vehicle has different components and operating parameters. In addition, different AI algorithms offer different predictive capabilities as well as varying levels of accuracy, depending on the predictive analytics provider used by the TSP.
As a result, fleets interested in adopting these solutions need to question their TSPs closely to determine whether their platform covers all vehicles in the fleet’s inventory and the accuracy of their predictions. Also check to see whether your provider’s weekly reports show too many expected malfunctions, indicating that their fault analysis is overly broad and likely to result in overservicing if maintenance teams attempt to address every red flag.
These issues will be resolved as more predictive analytics specialists enter the market to service TSPs wanting to add these capabilities. New features will likely be added as well, ranging from prescribing steps to avoid system or component failures to advising parts departments to order specific parts to correct a specific problem for each VIN number affected.
While today’s predictive maintenance solutions are focused exclusively on tractors, the technology will also eventually extend to trailers, door sensors, tire pressure, cargo and other aspects of fleet operations. These AI models will be leveraged to look for anomalies that can not only predict equipment failures but also improve asset monitoring in a variety of ways.
All of these solutions will help eliminate premature and emergency maintenance, reduce equipment maintenance costs, optimize field crew efficiencies, and – most importantly – keep trucks rolling. There is no such thing as a crystal ball, but predictive maintenance is a close equivalent for fleets.
Michael Bloom is Head of Marketing for Sensata INSIGHTS, a global business unit of Sensata Technologies that provides end-to-end IoT solutions spanning the entire supply chain including logistics, telematics, and worksite monitoring and management.
